Hyderabad : VISWAM.AI (A Joint initiative of Swecha and IIIT Hyderabad), jointly with SFLC.in released the new Draft License for consultation aimed to address critical gaps in current open Source licenses, regarding data usage, attribution, and the definition of openness in the age of AI. This License Draft Release was announced at the roundtable event “Understanding Trust and Safety in AI: From Code to Creativity” stakeholder consultation held at IIIT Hyderabad, jointly by VISWAM.AI, SFLC.in, FOSS United and The Linux Foundation.
Various community initiatives for the publication of datasets for public use have gained traction, exemplified by the community model pioneered by VISWAM.AI and Swecha for Telugu datasets. Nonetheless, the absence of suitable licensing frameworks presents a risk that these datasets could be appropriated by large organisations without appropriate attribution or reciprocal contribution to the community. The consultation highlighted that existing open-source licenses are insufficient for AI. While they cover source code, they fail to protect the training data which is the fuel of modern AI, from appropriation by proprietary giants.
The Key Objectives of the Draft License:
Community Ownership and Reciprocity: The license is designed to prevent the extraction of community-generated data by large corporations without attribution, with the central tenet to ensure that the value generated from the community-owned contributions flows back to the community. This is based on the principles of copy-left licenses like GNU GPL which are popular in the field of software.
Verifiability: It introduces the concept that for an AI to be truly open, the data provenance must be verifiable to detect bias and ensure safety. If the dataset is biased, the “openness” of the source code alone is insufficient.
In the following days, VISWAM.AI and SFLC.in will be conducting a series of online and in-person public consultations to gain feedback on the draft of the license. The draft license for consultation is available at: https://discuss.sflc.in/d/SBNXHa9k/viswam-ai-data-set-license-draft-for-discussion
“The Free and Open Source Software movement has been one of the most successful collaborative efforts in history. However, the licenses that powered this movement like the GPL and Creative Commons were written for a world of copying and distribution. They were designed to answer: “Can I copy this code?” They were never built to answer: “Can I train a neural network on this culture?
Even in the internet era, “to train or not to train” was a context the pre-AI world never envisaged. We are releasing this Draft License for consultation to ensure proprietary models cannot simply appropriate our work without attribution. This paves the way for the release of community-contributed, crowdsourced datasets under a framework that guarantees community ownership and verifiability” said Kiran Chandra Yarlagadda, Center Head and Chief Technologist, VISWAM.AI, Founder, Swecha.
Prasanth Sugathan, Legal Director, SFLC.IN said “We need a license to ensure that the rights of creators are protected and big tech companies do not appropriate the work of individuals and the community without even attributing them.”
